Any Warhol, Animal, Vegetable, Mineral is part of an ongoing series querying the hierarchy of classifications embedded in AI tools, in dialog with influences of contemporary art. These pictures were machine-generated with an initial image of Andy Warhol’s Self portrait (hair on end).
Visual object recognition software uses classification tags that are organized in a general taxonomy. For example, the visual database ImageNet contains millions of images sorted into thousands of categories. The prompts I used to generate the images explore top-level classifications like person, animal, plant, or sport, artifact, geological formation; and specific object categories like streetcar, tram, tramcar, trolley, trolley car.
Using the same initial image in all generations maintains visual consistency in color and composition across the entire series, foregrounding the specific terms used in the prompt by illustrating different concepts with the same shape. At the same time it pulls back the curtain on the knowledge systems (with all their assumptions and biases) that are built into the AI tools we use.
The use of Warhol’s self-portrait is a nod to the artist’s career-long practice of engaging in the “reproduction” of images rather than the “representation” of reality. I use AI to shift this to the generation of images rather than their reproduction while also relying on an entirely mechanical process.
Filed under: Machine Learning Text-to-Image Featured
Christopher Adams is an art producer and computer programmer based in Taipei.
